โกiApp ASR Base ๐
Benchmark results on Thai Testset from most common Dataset, Mozilla Common Voice. There are 2 version of testset that we are using to evaluate on iApp ASR Base.
Last updated
Benchmark results on Thai Testset from most common Dataset, Mozilla Common Voice. There are 2 version of testset that we are using to evaluate on iApp ASR Base.
Last updated
Test Set Size: 11,038 test examples
Average Word Error Rate: 1.1086199973229822
Average Word Error Rate (No Space): 0.663164235042163
Unseen dataset
Thai language only
Speaker: Male, Female, Children
https://docs.google.com/spreadsheets/d/1rh6nsw8qsZjDf3M8J9lepgnZHep5cD3rbJVOw-MTvj0/edit?usp=sharing
Test Set Size: 11,042 samples
Average Word Error Rate (WER) : 0.1452 or 14.52%
Average Character Error Rate (CER) : 0.587 or 5.87%
Average Accuracy based on WER : 85.48%
Average Accuracy based on CER : 94.13%
Response Time: 0.3 seconds (outperformed other vendors up to x15 times faster)
Unseen dataset
Thai language only
Speaker: Male, Female, Children